Skip to content
Images

DFlow2-Distill-4steps-eDMD-LADD-LoRA

Fast 4-step image generation

2 min readupdated

This model is a special LoRA (a small add-on for a base model to change its behavior) for the dflow_2b image generator by Data Archetype. The standard version requires 50 steps with two passes per step (100 total model passes), whereas this new release cuts the process down to just 4 steps with a single pass per step. Thanks to distillation based on the eDMD (distribution matching distillation) and LADD (texture critic) methods, the LoRA drastically accelerates image creation while maintaining acceptable quality and detail. The model is tailored for users of the popular ComfyUI interface running on NVIDIA graphics cards.

Key features

  • Fast process: Cuts standard generation down to 4 Euler steps instead of 50.
  • Special nodes: Includes custom nodes for ComfyUI to correctly apply the LoRA without weight merging.
  • Automatic size adjustment: Passes exact width, height, and self-attention gain parameters for different resolutions.
  • Broad training spectrum: Trained on short prompts, detailed descriptions, and layout JSON markup.

What you can do

  • Generate images: Use ready-made workflows in ComfyUI for rapid work based on text descriptions.
  • Use layouts: Apply structured JSON queries for precise element positioning.
  • Test different resolutions: Run generation on base proportions (for example, 1824x1024 or 1216x832).

Example

To run in ComfyUI, download the base model dflow_2b, add the required nodes, and place the LoRA file in the ComfyUI/models/loras/ folder:

cd ComfyUI/custom_nodes
hf download data-archetype/dflow_2b --local-dir dflow_2b
pip install timm
Read moreDetails
Base model: data-archetype/dflow_2b
Architecture: Rank 64 on attention and MLP layers (30 blocks total)
Requirements: NVIDIA GPU with CUDA support (does not work on Mac with MPS)
License: MG-BY-NC 2.0 (non-commercial)
Examples
Discuss Telegram 2 29

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 4.2 GB · ready files6.2 GB
  • FP8 / INT8recommendedweights 2.3 GB4.0 GB
  • 4-bitweights 1.4 GB3.0 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 5060 Ti 8GB

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model dflow_2b.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # lvladikov/DFlow2-Distill-4steps-eDMD-LADD-LoRA
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download data-archetype/dflow_2b --local-dir ~/models/dflow_2b
    hf download lvladikov/DFlow2-Distill-4steps-eDMD-LADD-LoRA --local-dir ~/models/DFlow2-Distill-4steps-eDMD-LADD-LoRA
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to dflow_2b
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • Nunchaku (SVDQuant 4-bit)v1.2.1from January 25, 2026

    ≈3× faster and 3.5× less memory

    4-bit versions of popular image models with almost no quality loss (FP4 on RTX 50).

    How to enable: Ready weights for supported models; in ComfyUI use ComfyUI-nunchaku.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

Requires the base model

data-archetype/dflow_2b

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. t5gemma-2-270m-270mfamily root
  2. dflow_2b
  3. DFlow2-Distill-4steps-eDMD-LADD-LoRA

New posts about versions of t5gemma-2-270m-270m (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face
LoRAs and adapters 3All on Hugging Face

About the model

lvladikov/DFlow2-Distill-4steps-eDMD-LADD-LoRA
Author
lvladikov
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

292

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments